4 * Copyright(c) 2010-2012 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * * Neither the name of Intel Corporation nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
40 #include <sys/types.h>
42 #include <sys/queue.h>
43 #include <netinet/in.h>
50 #include <rte_common.h>
52 #include <rte_memory.h>
53 #include <rte_memcpy.h>
54 #include <rte_memzone.h>
55 #include <rte_tailq.h>
57 #include <rte_per_lcore.h>
58 #include <rte_launch.h>
59 #include <rte_atomic.h>
60 #include <rte_cycles.h>
61 #include <rte_prefetch.h>
62 #include <rte_lcore.h>
63 #include <rte_per_lcore.h>
64 #include <rte_branch_prediction.h>
65 #include <rte_interrupts.h>
67 #include <rte_random.h>
68 #include <rte_debug.h>
69 #include <rte_ether.h>
70 #include <rte_ethdev.h>
72 #include <rte_mempool.h>
77 #define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
79 #define L2FWD_MAX_PORTS 32
81 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
85 * RX and TX Prefetch, Host, and Write-back threshold values should be
86 * carefully set for optimal performance. Consult the network
87 * controller's datasheet and supporting DPDK documentation for guidance
88 * on how these parameters should be set.
90 #define RX_PTHRESH 8 /**< Default values of RX prefetch threshold reg. */
91 #define RX_HTHRESH 8 /**< Default values of RX host threshold reg. */
92 #define RX_WTHRESH 4 /**< Default values of RX write-back threshold reg. */
95 * These default values are optimized for use with the Intel(R) 82599 10 GbE
96 * Controller and the DPDK ixgbe PMD. Consider using other values for other
97 * network controllers and/or network drivers.
99 #define TX_PTHRESH 36 /**< Default values of TX prefetch threshold reg. */
100 #define TX_HTHRESH 0 /**< Default values of TX host threshold reg. */
101 #define TX_WTHRESH 0 /**< Default values of TX write-back threshold reg. */
103 #define MAX_PKT_BURST 32
104 #define BURST_TX_DRAIN 200000ULL /* around 100us at 2 Ghz */
109 * Configurable number of RX/TX ring descriptors
111 #define RTE_TEST_RX_DESC_DEFAULT 128
112 #define RTE_TEST_TX_DESC_DEFAULT 512
113 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
114 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
116 /* ethernet addresses of ports */
117 static struct ether_addr l2fwd_ports_eth_addr[L2FWD_MAX_PORTS];
119 /* mask of enabled ports */
120 static uint32_t l2fwd_enabled_port_mask = 0;
122 /* list of enabled ports */
123 static uint32_t l2fwd_dst_ports[L2FWD_MAX_PORTS];
125 static unsigned int l2fwd_rx_queue_per_lcore = 1;
127 #define MAX_PKT_BURST 32
130 struct rte_mbuf *m_table[MAX_PKT_BURST];
133 #define MAX_RX_QUEUE_PER_LCORE 16
135 /* Each VF(port) has one Rx/Tx queue (with queueid: 0) */
136 struct lcore_queue_conf {
139 unsigned rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
140 unsigned tx_queue_id;
141 struct mbuf_table tx_mbufs[L2FWD_MAX_PORTS];
143 } __rte_cache_aligned;
144 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
146 static const struct rte_eth_conf port_conf = {
149 .header_split = 0, /**< Header Split disabled */
150 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
151 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
152 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
153 .hw_strip_crc = 1, /**< CRC stripped by hardware */
159 static const struct rte_eth_rxconf rx_conf = {
161 .pthresh = RX_PTHRESH,
162 .hthresh = RX_HTHRESH,
163 .wthresh = RX_WTHRESH,
167 static const struct rte_eth_txconf tx_conf = {
169 .pthresh = TX_PTHRESH,
170 .hthresh = TX_HTHRESH,
171 .wthresh = TX_WTHRESH,
173 .tx_free_thresh = 0, /* Use PMD default values */
174 .tx_rs_thresh = 0, /* Use PMD default values */
177 struct rte_mempool * l2fwd_pktmbuf_pool = NULL;
179 /* Per-port statistics struct */
180 struct l2fwd_port_statistics {
184 } __rte_cache_aligned;
185 struct l2fwd_port_statistics port_statistics[L2FWD_MAX_PORTS];
187 /* A tsc-based timer responsible for triggering statistics printout */
188 #define TIMER_MILLISECOND 2000000ULL /* around 1ms at 2 Ghz */
189 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
190 static int64_t timer_period = 10 * TIMER_MILLISECOND * 1000; /* default period is 10 seconds */
192 /* Print out statistics on packets dropped */
196 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
199 total_packets_dropped = 0;
200 total_packets_tx = 0;
201 total_packets_rx = 0;
203 const char clr[] = { 27, '[', '2', 'J', '\0' };
204 const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
206 /* Clear screen and move to top left */
207 printf("%s%s", clr, topLeft);
209 printf("\nPort statistics ====================================");
211 for (portid = 0; portid < L2FWD_MAX_PORTS; portid++) {
212 /* skip ports that are not enabled */
213 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
216 printf("\nStatistics for port %u ------------------------------"
217 "\nPackets sent: %24"PRIu64
218 "\nPackets received: %20"PRIu64
219 "\nPackets dropped: %21"PRIu64,
221 port_statistics[portid].tx,
222 port_statistics[portid].rx,
223 port_statistics[portid].dropped);
225 total_packets_dropped += port_statistics[portid].dropped;
226 total_packets_tx += port_statistics[portid].tx;
227 total_packets_rx += port_statistics[portid].rx;
229 printf("\nAggregate statistics ==============================="
230 "\nTotal packets sent: %18"PRIu64
231 "\nTotal packets received: %14"PRIu64
232 "\nTotal packets dropped: %15"PRIu64,
235 total_packets_dropped);
236 printf("\n====================================================\n");
239 /* Send the packet on an output interface */
241 l2fwd_send_burst(struct lcore_queue_conf *qconf, unsigned n, uint8_t port)
243 struct rte_mbuf **m_table;
247 queueid = (uint16_t) qconf->tx_queue_id;
248 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
250 ret = rte_eth_tx_burst(port, (uint16_t) queueid, m_table, (uint16_t) n);
252 port_statistics[port].tx += ret;
253 if (unlikely(ret < n)) {
254 port_statistics[port].dropped += (n - ret);
256 rte_pktmbuf_free(m_table[ret]);
263 /* Send the packet on an output interface */
265 l2fwd_send_packet(struct rte_mbuf *m, uint8_t port)
267 unsigned lcore_id, len;
268 struct lcore_queue_conf *qconf;
270 lcore_id = rte_lcore_id();
272 qconf = &lcore_queue_conf[lcore_id];
273 len = qconf->tx_mbufs[port].len;
274 qconf->tx_mbufs[port].m_table[len] = m;
277 /* enough pkts to be sent */
278 if (unlikely(len == MAX_PKT_BURST)) {
279 l2fwd_send_burst(qconf, MAX_PKT_BURST, port);
283 qconf->tx_mbufs[port].len = len;
288 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
290 struct ether_hdr *eth;
294 dst_port = l2fwd_dst_ports[portid];
295 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
297 /* 00:09:c0:00:00:xx */
298 tmp = ð->d_addr.addr_bytes[0];
299 *((uint64_t *)tmp) = 0x000000c00900 + (dst_port << 24);
302 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
304 l2fwd_send_packet(m, (uint8_t) dst_port);
307 /* main processing loop */
309 l2fwd_main_loop(void)
311 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
314 unsigned int nb_ports;
315 uint64_t prev_tsc = 0;
316 uint64_t diff_tsc, cur_tsc, timer_tsc;
317 unsigned i, j, portid, nb_rx;
318 struct lcore_queue_conf *qconf;
322 lcore_id = rte_lcore_id();
323 qconf = &lcore_queue_conf[lcore_id];
325 nb_ports = rte_eth_dev_count();
326 if (nb_ports > L2FWD_MAX_PORTS)
327 nb_ports = L2FWD_MAX_PORTS;
329 if (qconf->n_rx_queue == 0) {
330 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
334 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
336 for (i = 0; i < qconf->n_rx_queue; i++) {
338 portid = qconf->rx_queue_list[i];
339 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
345 cur_tsc = rte_rdtsc();
348 * TX burst queue drain
350 diff_tsc = cur_tsc - prev_tsc;
351 if (unlikely(diff_tsc > BURST_TX_DRAIN)) {
353 for (portid = 0; portid < nb_ports; portid++) {
354 if (qconf->tx_mbufs[portid].len == 0)
356 l2fwd_send_burst(&lcore_queue_conf[lcore_id],
357 qconf->tx_mbufs[portid].len,
359 qconf->tx_mbufs[portid].len = 0;
362 /* if timer is enabled */
363 if (timer_period > 0) {
365 /* advance the timer */
366 timer_tsc += diff_tsc;
368 /* if timer has reached its timeout */
369 if (unlikely(timer_tsc >= (uint64_t) timer_period)) {
371 /* do this only on master core */
372 if (lcore_id == rte_get_master_lcore()) {
374 /* reset the timer */
383 * Read packet from RX queues
385 for (i = 0; i < qconf->n_rx_queue; i++) {
387 portid = qconf->rx_queue_list[i];
388 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
389 pkts_burst, MAX_PKT_BURST);
391 port_statistics[portid].rx += nb_rx;
393 for (j = 0; j < nb_rx; j++) {
395 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
396 l2fwd_simple_forward(m, portid);
403 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
411 l2fwd_usage(const char *prgname)
413 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
414 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
415 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
416 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
421 l2fwd_parse_portmask(const char *portmask)
426 /* parse hexadecimal string */
427 pm = strtoul(portmask, &end, 16);
428 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
438 l2fwd_parse_nqueue(const char *q_arg)
443 /* parse hexadecimal string */
444 n = strtoul(q_arg, &end, 10);
445 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
449 if (n >= MAX_RX_QUEUE_PER_LCORE)
456 l2fwd_parse_timer_period(const char *q_arg)
461 /* parse number string */
462 n = strtol(q_arg, &end, 10);
463 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
465 if (n >= MAX_TIMER_PERIOD)
471 /* Parse the argument given in the command line of the application */
473 l2fwd_parse_args(int argc, char **argv)
478 char *prgname = argv[0];
479 static struct option lgopts[] = {
485 while ((opt = getopt_long(argc, argvopt, "p:q:T:",
486 lgopts, &option_index)) != EOF) {
491 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
492 if (l2fwd_enabled_port_mask == 0) {
493 printf("invalid portmask\n");
494 l2fwd_usage(prgname);
501 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
502 if (l2fwd_rx_queue_per_lcore == 0) {
503 printf("invalid queue number\n");
504 l2fwd_usage(prgname);
511 timer_period = l2fwd_parse_timer_period(optarg) * 1000 * TIMER_MILLISECOND;
512 if (timer_period < 0) {
513 printf("invalid timer period\n");
514 l2fwd_usage(prgname);
521 l2fwd_usage(prgname);
525 l2fwd_usage(prgname);
531 argv[optind-1] = prgname;
534 optind = 0; /* reset getopt lib */
539 MAIN(int argc, char **argv)
541 struct lcore_queue_conf *qconf;
543 unsigned int nb_ports;
545 unsigned lcore_id, rx_lcore_id;
547 unsigned nb_ports_in_mask = 0;
550 ret = rte_eal_init(argc, argv);
552 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
556 /* parse application arguments (after the EAL ones) */
557 ret = l2fwd_parse_args(argc, argv);
559 rte_exit(EXIT_FAILURE, "Invalid L2FWD-VF parameters\n");
561 /* create the mbuf pool */
563 rte_mempool_create("mbuf_pool", NB_MBUF,
565 sizeof(struct rte_pktmbuf_pool_private),
566 rte_pktmbuf_pool_init, NULL,
567 rte_pktmbuf_init, NULL,
569 if (l2fwd_pktmbuf_pool == NULL)
570 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
573 #ifdef RTE_LIBRTE_IGB_PMD
574 if (rte_igb_pmd_init() < 0)
575 rte_exit(EXIT_FAILURE, "Cannot init igb pmd\n");
577 #ifdef RTE_LIBRTE_IXGBE_PMD
578 if (rte_ixgbe_pmd_init() < 0)
579 rte_exit(EXIT_FAILURE, "Cannot init ixgbe pmd\n");
581 if (rte_ixgbevf_pmd_init() < 0)
582 rte_exit(EXIT_FAILURE, "Cannot init ixgbevf pmd\n");
585 if (rte_eal_pci_probe() < 0)
586 rte_exit(EXIT_FAILURE, "Cannot probe PCI\n");
588 nb_ports = rte_eth_dev_count();
590 rte_exit(EXIT_FAILURE, "No Ethernet port - bye\n");
592 if (nb_ports > L2FWD_MAX_PORTS)
593 nb_ports = L2FWD_MAX_PORTS;
595 /* reset l2fwd_dst_ports */
596 for (portid = 0; portid < L2FWD_MAX_PORTS; portid++)
597 l2fwd_dst_ports[portid] = 0;
601 qconf = &lcore_queue_conf[rx_lcore_id];
604 * Initialize the lcore/port-rx-queue configuration of each lcore.
605 * NOTE: Each logical core sends packets out to all port-tx-queues
607 for (portid = 0; portid < nb_ports; portid++) {
609 /* skip ports that are not enabled */
610 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
613 if (nb_ports_in_mask % 2) {
614 l2fwd_dst_ports[portid] = last_port;
615 l2fwd_dst_ports[last_port] = portid;
622 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
623 lcore_queue_conf[rx_lcore_id].n_rx_queue ==
624 l2fwd_rx_queue_per_lcore) {
627 if (rx_lcore_id >= RTE_MAX_LCORE)
628 rte_exit(EXIT_FAILURE, "Not enough cores\n");
631 qconf = &lcore_queue_conf[rx_lcore_id];
632 qconf->tx_queue_id = 0;
633 qconf->rx_queue_list[qconf->n_rx_queue] = portid;
636 printf("Lcore %u: RX port %u\n", rx_lcore_id, portid);
640 /* Initialise each port */
641 for (portid = 0; portid < nb_ports; portid++) {
643 /* skip ports that are not enabled */
644 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
645 printf("Skipping disabled port %u\n", portid);
650 printf("Initializing port %u... ", portid);
652 ret = rte_eth_dev_configure((uint8_t) portid, 1, 1, &port_conf);
654 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
657 rte_eth_macaddr_get((uint8_t) portid,
658 &l2fwd_ports_eth_addr[portid]);
660 /* init one RX queue */
662 ret = rte_eth_rx_queue_setup((uint8_t) portid, 0, nb_rxd,
666 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, port=%u\n",
669 /* init one TX queue */
671 ret = rte_eth_tx_queue_setup((uint8_t) portid, 0, nb_txd,
674 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, port=%u\n",
678 ret = rte_eth_dev_start((uint8_t) portid);
680 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%u\n",
686 printf("Port %u, MAC address: %02X:%02X:%02X:%02X:%02X:%02X\n\n",
688 l2fwd_ports_eth_addr[portid].addr_bytes[0],
689 l2fwd_ports_eth_addr[portid].addr_bytes[1],
690 l2fwd_ports_eth_addr[portid].addr_bytes[2],
691 l2fwd_ports_eth_addr[portid].addr_bytes[3],
692 l2fwd_ports_eth_addr[portid].addr_bytes[4],
693 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
695 /* initialize port stats */
696 memset(&port_statistics, 0, sizeof(port_statistics));
699 /* launch per-lcore init on every lcore */
700 rte_eal_mp_remote_launch(l2fwd_launch_one_lcore, NULL, CALL_MASTER);
701 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
702 if (rte_eal_wait_lcore(lcore_id) < 0)